soul-memory

What This Skill Does

The Soul Memory skill is an advanced long-term memory management framework specifically engineered for AI agents operating within the OpenClaw ecosystem. Moving beyond standard context windows, this system uses 8 specialized modules to parse, classify, and store agent interactions. By implementing heartbeat deduplication (via MD5 hashing), it ensures that your agent does not bloat its memory with redundant data. The integration with the OpenClaw Plugin v0.2.1-beta allows for automatic context injection, meaning your agent gains 'permanent' knowledge of user preferences, past conversations, and specific domain facts, effectively creating a persistent persona across multiple sessions.

Installation

To integrate Soul Memory into your OpenClaw environment, you can use the automated installation flow provided by the OpenClaw hub. Run the following command in your terminal: clawhub install openclaw/skills/skills/kingofqin2026/soul-memory. Alternatively, for manual control, clone the repository from GitHub and execute bash install.sh. After the installation completes, ensure you run openclaw gateway restart to allow the plugin to hook into the system's before_prompt_build process, which triggers the intelligent context injection before the agent generates a response.

Use Cases

• Personalized AI Assistants: Maintain a history of user likes, dislikes, and technical preferences to ensure the agent feels 'aware' of the user over months of interaction.
• Project Management: Store specific QST physics data or development requirements that the AI needs to reference across multiple coding sessions.
• Multilingual Support: Utilize the Cantonese Branch for nuanced language handling and context-aware responses, perfect for regional localization.
• System Diagnostics: Use the CLI interface to query memory stats or clean up outdated entries using the version control rollback module.

Example Prompts

1. "Soul Memory, recall what I told you about my preferred CSS framework settings yesterday."
2. "Add a memory: [I] The project deadline has been moved to next Friday at 5 PM."
3. "Search for all physics-related notes in my database and summarize the key findings for the QST module."

Tips & Limitations

• Lenient Mode: Enable this in the config if you find the agent is missing subtle information; however, note that this increases storage footprint.
• Versioning: Always utilize the D module version control before performing bulk updates to ensure you can roll back in case of data corruption.
• Performance: While the system is highly optimized, indexing thousands of memories may increase initial boot time for the pre_response_trigger. Keep your memory tags organized using the Priority Parser to maintain performance.